Incidental urinary tract injury and the formation of vesicovaginal fistula at the time of hysterectomy for benign indications.

A total of 837,737 hysterectomies were performed between 2005 and 2011. After excluding 177,283 hysterectomies with insufficient information regarding laparoscopic use and 79,059 with malignancy, complete information was available on 581,395 individuals ( Table 1 ). Average age of patients was 47.8 years and approximately 60% identified as white. Most procedures were performed inpatient (86.5%) and were paid for by private insurance (93.9%). Thirty-three percent of procedures utilized laparoscopic assistance ( Table 1 ). Urinary tract injuries were reported in 1.15% of patients and VVF in 0.11%. Of those that had a post-hysterectomy fistula repair, 20.5% had an antecedent urinary tract injury at the time of the index hysterectomy, while among subjects with identified urinary tract injury, 2.0% had a fistula repair ( Table 2 ). Report of injury to the urinary tract in any type of hysterectomy was associated with 20-fold increased odds of undergoing VVF (OR 20.64; 95% CI 16.98, 25.10) ( Table 2 ). Compared to women without detected intraoperative injuries, those with detected injuries were 30-fold more likely to undergo VVF repair if their surgery involved laparoscopy but only 17-fold more likely if their surgery did not involve any use of laparoscopy ( Table 2 ). We detected a statistically significant interaction between laparoscopy and injury in relation to VVF repair irrespective of hysterectomy type (p=0.014). After detecting a statistical interaction between recognized injury and laparoscopy, we evaluated if laparoscopy alone was independently associated with undergoing fistula repair. In multivariable-adjusted models including injuries, women who had a procedure involving laparoscopy were 1.53 times (OR 1.53; 95% CI 1.26, 1.87) more likely to have a VVF repair than those without laparoscopy ( Table 4 ). We repeated this assessment in individuals with no reported injury at the time of the hysterectomy and the OR remained positive (OR 1.40; 95% CI 1.12, 1.75) ( Table 4 ). We then evaluated the association between laparoscopy and VVF repair across hysterectomy categories (total, sub-total and vaginal) ( Table 4 ). Laparoscopy was not uniformly associated with increased risk in all hysterectomy categories (p<0.001). Compared to those who had a TAH, TLH was not associated with increased risk of VVF repair in models including injury (OR 0.94) and excluding injury (OR 0.89). LSH was associated with 2.5-fold higher odds (OR 2.47; 95% CI 1.25, 4.86) of having VVF than SAH. The odds ratio was even higher for LAVH (OR 2.82; 95% CI 1.98, 4.02). Magnitudes of association were similar when excluding women with injury from analysis ( Table 4 ). To identify factors associated with fistula repair other than injury, we evaluated the association between types of hysterectomy procedures ( Tables 3 – 4 ). In multivariable adjusted models including injury, compared to individuals having a TAH, those having had a SAH (OR 0.13; 95% CI 0.08, 0.22), LSH (OR 0.27; 95% CI 0.18, 0.32), TVH (OR 0.23; 95% CI 0.17, 0.32) and LAVH (OR 0.66; 95% CI 0.52, 0.83) had lower odds of having a fistula ( Table 3 ). Re-evaluating this association limited to individuals without reported injuries did not change observed associations appreciably ( Table 3 ). Finally, we evaluated the risk of recording a urinary tract injury by hysterectomy types. In multivariable adjusted models, compared to TAH, all other surgical types were associated with reduced odds of injury, with lowest odds for LSH (OR 0.42; 95% CI 0.37, 0.48) ( Table 5 ). Overall, the laparoscopic route was associated with lower odds of injury reports (OR 0.59; 95% CI 0.52, 0.67) than non-laparoscopic procedures. The association observed for laparoscopy and injury differed by hysterectomy category (p<0.001). TLH was less likely to have injury (OR 0.59; 95% CI 0.52, 0.68) than TAH, and injury was also less likely in LSH (OR 0.54; 95% CI 0.46, 0.63) than SAH. However, among those undergoing vaginal hysterectomy, LAVH had similar odds of injury as TVH (OR 0.98; 95% CI 0.90, 1.07).

We performed a retrospective cohort study utilizing the Healthcare Cost and Utilization Project (HCUP) databases, specifically the State Inpatient Databases (SID) and The State Ambulatory Surgery and Services Databases (SASD). HCUP includes the largest collection of longitudinal hospital care data in the United States, with all-payer, encounter-level information. Details regarding HCUP are described elsewhere [ 13 ]. The SID and SASD have supplemental revisit information, which allowed patient-level longitudinal tracking. We obtained SID, SASD, and associated supplemental revisit information for California, New York, and Florida from 2005–2011. We combined SID and SASD data to include all recorded hysterectomies in the selected states. Hysterectomies were identified in women aged 18 and above. SID contains ICD-9 procedural codes, while SASD uses CPT codes; therefore, the combination of both databases was used to define hysterectomy, urinary tract repair, and fistula repair procedures ( Appendix A ). VVF repair codes were identified within 1 year after hysterectomy up to 2011. The Vanderbilt University Medical Center Institutional Review Board deemed this study to be exempt. The primary outcomes for this study were reported urinary tract injury during hysterectomy and subsequent repair of VVF. We identified three main types of hysterectomy: 1) total abdominal hysterectomy, 2) subtotal abdominal hysterectomy and 3) vaginal hysterectomy. The hysterectomies were then further subtyped by laparoscopy use as follows: total abdominal hysterectomy (TAH- referent category), total laparoscopic hysterectomy (TLH), subtotal (supracervical) abdominal hysterectomy (SAH), laparoscopic subtotal (supracervical) hysterectomy (LSH), total vaginal hysterectomy (TVH) and laparoscopic-assisted vaginal hysterectomy (LAVH). We excluded hysterectomies performed for cancer or if a gynecologic cancer diagnostic code was noted within one month of the surgery ( Appendix B ). Procedures with unclear use of laparoscopy were excluded. Information on robotic assistance was not available, therefore robotic versus straight-stick laparoscopy cannot be delineated. Information regarding any previous radiotherapy and previous pelvic surgery were not available through the HCUP datasets. Participant characteristics are reported in Table 1 . To evaluate all reported associations, we constructed a multivariable adjusted logistic regression model (base multivariable model), which includes age at procedure, diabetes, hypertension, obesity status, race, medical insurance type, state, urban location of medical center, inpatient vs. outpatient surgery, hysterectomy category (total, subtotal, and vaginal), and laparoscopy use (yes/no). We evaluated the association between report of urinary tract injury and subsequent VVF repair using the base multivariable model. We then evaluated whether use of laparoscopy during hysterectomy modifies the association between injury and VVF by performing interaction analysis. To further understand which hysterectomy subtypes were associated with the greatest risk of VVF we present multivariable adjusted results for all six subtypes of hysterectomy using TAH as the referent procedure. To examine the independent contribution of hysterectomy type on VVF we performed the analysis 1) in all eligible hysterectomy patients including individuals for whom an injury was reported (while also adjusting for it) and 2) excluding individuals with injuries reported during hysterectomy. To understand if laparoscopy was a risk factor for VVF, independent of the hysterectomy category and report of injury, we first evaluated the association between laparoscopic assistance and VVF using the base multivariable model. We then used models stratified by hysterectomy subtype to understand if the association between laparoscopy and VVF differed by hysterectomy subtype. To further evaluate whether this association is driven through reports of injury, we performed the same analyses excluding individuals for whom injury was reported during hysterectomy. Finally, we evaluated which of the six hysterectomy types is associated with the highest risk for VVF repair, whether laparoscopy was an independent risk factor for VVF repair, and whether the association between laparoscopy and VVF repair differed by hysterectomy type in multivariable adjusted models. All formal interaction analyses were computed with the likelihood ratio test. All statistical analyses are performed using STATA 16.0 (College Station, TX) in accordance with reporting guidelines established by HCUP.

Our study demonstrates the magnitude of association between laparoscopy and the need for subsequent VVF repair. It also highlights that association between injury and VVF repair varies by hysterectomy type. We confirm the association between injury identification at the time of hysterectomy and subsequent VVF repair, but also note approximately 80% of the VVF cases in our population were found in individuals in whom no injury was reported during hysterectomy. A strength of our study is that the outcomes we assessed appear to be congruent with previously reported populations. The overall rate of urinary tract injury in our cohort is 1.2%. This concurs with a meta-analysis of 79 studies that reported the overall rate of urinary tract injury at the time of benign hysterectomy around 0.8% [ 14 ], and with a study based out of California that reported genitourinary tract injury to be about 1.8% (a subpopulation in our study) [ 8 ]. Another strength of our study is utilizing the HCUP database, which allowed us to conduct a large population-based study, representing different types of practices in multiple geographic regions in the United States, and capturing both inpatient and outpatient procedures. Our study was less dependent on return to the same facility, as any post-operative complications that were followed up in the same state as the original procedure was captured. Finally, the large number of urinary tract injuries and VVF events captured allowed us to assess relationships between hysterectomy type, injury and VVF in greater depth than previously published studies. However, as with most administrative databases, there are inherent limitations. One such limit is that our data are dependent on the accuracy of the coding performed. Related to coding is the inability to further delineated types of laparoscopies. Since the use of robotic-assisted laparoscopy was poorly defined, we cannot apply our findings to a technology which is now being more commonly applied. We also must consider how the learning curve associated with the application of technology which was new at the time of our data collection affects the outcomes reported. There is also no standard definition of how much laparoscopy constitutes a laparoscopic-assisted case, so our findings cannot be broadly applied to all laparoscopies, particularly LAVH. Another limitation of our study is that we are unable to account for cases that were converted to a different approach, and instances where a conversion is necessary could increase the risk of intraoperative injury or fistula formation. Working within the scope of the dataset, we excluded individuals with recent diagnosis of cancer or gynecologic cancer, however information regarding past pelvic radiation, past pelvic surgery, or specific indication for the hysterectomy could not be obtained. Previous therapies or certain indications for surgery can significantly increase the risk of operative complications, and our findings should be considered in light of the possibility this could in part explain findings from this study if these factors were also associated with choice of route and type of hysterectomy. Furthermore, post-operative complications can also contribute to VVF formation, however, due to lack of this information, we are not able to assess the potential associations between hysterectomy choice and complications and the potential mediating role of these complications on VVF. Our study improves the existing literature by quantifying our findings by subtype of hysterectomy. Three other studies have evaluated the relationship between hysterectomy type, genitourinary injury and fistula formation in large numbers [ 15 , 6 , 8 ]. Hesselman et. al reported a 15-fold increased risk of pelvic organ fistula formation associated with organ injury (OR = 15.3) and reported a higher risk of fistula formation associated with total abdominal (OR = 2.25) and laparoscopic hysterectomy (OR = 1.75) compared to vaginal hysterectomy (concomitant procedures not discussed) [ 15 ]. Although these findings are not directly comparable to our study, as they included vascular and bowel injuries in addition to urinary tract injuries and included all pelvic organ fistulas, some of our findings are in agreement, including the lower odds of VVF in TVH (OR = 0.22) and LAVH (OR = 0.66) compared with TAH [ 15 ]. In another study Forsgren et. al did not evaluate injury rates by hysterectomy subtype, but similar to our study, reported higher rate of fistula repair surgery (82.2 per 100,000 person-years) among patients undergoing laparoscopic hysterectomy (including TLH and LAVH) compared to abdominal (total 12.9, subtotal 1.59) and vaginal (9.18) surgery [ 6 ]. The third study by Dallas et. al, reported lower risk of injury associated with laparoscopy use (OR = 0.59) compared with total abdominal surgery. The authors also reported higher risk of fistula formation if injury was detected on the same day than if it was delayed in detection, however, did not evaluate the relationship between hysterectomy subtype and fistula formation [ 8 ]. With regards to urinary tract injury, in contrast to our findings, a 2015 Cochrane review showed increased association of bladder or ureteral injury (together but not separately) when doing hysterectomy laparoscopically compared to abdominally [ 16 ]. This Cochrane review included randomized controlled trials and the search for injuries/outcomes may have been more thorough for laparoscopic procedures in these settings than for surgeries performed in non-study settings. Findings from our study bridge these seemingly disparate findings. Since injury during surgery is suspected to be a main cause of post-operative fistula formation, our observation of lower rates of injury but higher rates of VVF repair suggests the injury rate may be higher than what is detected and reported for laparoscopic procedures, or that the detection is better for non-laparoscopic procedures in observational study/real world clinical settings. Our observation that most VVF repairs were in individuals without reported injury further supports this idea. We postulate the association between using laparoscopy and subsequent repair of fistula is possibly due to the nature of when laparoscopic assistance is employed. Typically, it is thought that laparoscopy can provide a helpful adjunct in cases with a large uterine size, endometriosis, presence of fibroids, presence of scar tissue, previous cesarean delivery, or other abdominal surgeries. These same factors increase the risk of urinary tract injury and make it difficult to detect injury. There could be confounding given the adoption rate of laparoscopic hysterectomy, varying surgical case numbers and time to achieve proficiency in laparoscopy, and evolution of surgical equipment. This work informs how type of hysterectomy contributes to risk of fistula repair, how urinary tract injury at the time of hysterectomy contributes to that risk, and how laparoscopy can be variably associated with injury and subsequent fistula repair, depending on the subtypes of hysterectomy being performed. Future work would ideally compare the rates of injury and fistula repair between traditional laparoscopy and robotic-assisted laparoscopy, as well as try to incorporate relevant information regarding past surgical history pre-operative therapies, indications for benign hysterectomy, and other post-operative complications to assess the confounding and mediating effects of these factors on the associations between hysterectomy type, injury and fistula repair.

Vesicovaginal fistula (VVF), an abnormal channel between the bladder and the vagina, results in continuous leakage of urine from the vagina. Incidence of VVF after benign hysterectomy ranges from 0.1% to 0.28%, yet hysterectomy accounts for up to 82% of VVF cases, suggesting injury to the urinary tract during hysterectomy as the main cause [ 1 – 11 ]. A systematic review examining the incidence of urinary tract injuries associated with laparoscopic hysterectomy found 3.4% of injuries resulted in a VVF [ 12 ]. Previous studies assess risk associated between laparoscopy use and injury [ 8 ] and risk associated between laparoscopy use and VVF [ 6 ] separately. Collectively, these studies report paradoxical findings, where laparoscopy is associated with lower risk of injury during hysterectomy and yet higher risk of VVF. However, to our knowledge, large studies have not assessed the relationship between hysterectomy type, injury and VVF together in the same population. Knowing these relationships could aid the surgeon in choosing the route of surgery, could improve patient counseling, and could possibly encourage surgeons to use more caution when evaluating for intraoperative injuries after certain types of hysterectomies. We postulate that laparoscopy increases the risk of fistula repair, and that overall rates of VVF differ by type of hysterectomy. We designed a study to examine the complex relationships between urinary tract injury, type of hysterectomy and use of laparoscopy.